Improving Contextual Faithfulness of Large Language Models via Retrieval Heads-Induced Optimization
This addresses the issue of building trustworthy information-seeking systems in long-form question-answering, representing a novel method for a known bottleneck.
The paper tackled the problem of ensuring contextual faithfulness in retrieval-augmented large language models for long-form question-answering by proposing RHIO, a framework that improves faithfulness by teaching models to discriminate between faithful and unfaithful generations, resulting in significant improvements that outperform GPT-4o on the introduced GroundBench benchmark.
Ensuring contextual faithfulness in retrieval-augmented large language models (LLMs) is crucial for building trustworthy information-seeking systems, particularly in long-form question-answering (LFQA) scenarios. In this work, we identify a salient correlation between LFQA faithfulness and retrieval heads, a set of attention heads responsible for retrieving contextual information. Leveraging this insight, we propose RHIO, a framework designed to teach LLMs to explicitly discriminate between faithful and unfaithful generations. RHIO first augments unfaithful samples that simulate realistic model-intrinsic errors by selectively masking retrieval heads. Then, these samples are incorporated into joint training, enabling the model to distinguish unfaithful outputs from faithful ones conditioned on control tokens. Furthermore, these control tokens are leveraged to self-induce contrastive outputs, amplifying their difference through contrastive decoding. Additionally, to facilitate the evaluation of contextual faithfulness, we also introduce GroundBench, a comprehensive benchmark compiled from five existing LFQA datasets. Extensive experimental results on GroundBench demonstrate that RHIO significantly improves faithfulness, even outperforming GPT-4o.